[R] Diebold Mariano

Brajkovic J. Jurica.Brajkovic at soton.ac.uk
Fri Jun 25 18:07:53 CEST 2010


Hello, 
I am trying to calculate Diebold Mariano test statistic (DM) using dm.test module. I also try to do the same thing with STATA and I get vastly different results (4.5 vs 25). Does someone have experience with this module? 

I tried to calculate the DM statistic manually. If by “d” I define the difference of squared forecast errors for two models, then DM=mean(d)/sqrt(long_run_var(d)). To calculate long run variance of d I use newey west standard errors. What I don’t understand in newey west command is meaning of “prewhite”. Any help? 

Thanks!



More information about the R-help mailing list